CSV header validation ignores `ACCEPT_CASE_INSENSITIVE_PROPERTIES`: add `CsvReadFeature.CASE_INSENSITIVE_HEADERS`

[polyoxidonium]

After upgrading to Jackson 3.1.x, CSV parsing with schema-based deserialization
(schemaFor(...).withHeader().withColumnReordering(true)) fails when header
names differ only by case.

Example:

• expected: temp_max
• actual: TEMP_MAX

Error:
CsvReadException: Missing header column: ["temp_max"]

Mapper is configured with:
.enable(MapperFeature.ACCEPT_CASE_INSENSITIVE_PROPERTIES)

Expected behavior:
Case-insensitive property matching should also apply during header validation,
so headers differing only in case should not be treated as missing.

Actual behavior:
Header validation is strictly case-sensitive and fails before databind phase.

Notes:

• This behavior became visible after PR Check for missed columns even if given and expected column count are the same #580, which introduced stricter header validation.
• The issue is particularly visible when using Kotlin data classes and schemaFor(...),
  where expected column names are derived from property names.

Question:
Is this intended behavior, or should case-insensitive matching be applied during
header/schema reconciliation as well?
We have a lot of parsers in place which are supposed to work also with case-insensitive headers, so this can cause a lot of production issues on our side.

[cowtowncoder]

This is an oversight.

But there is a problem: MapperFeature is databind-level concept, whereas CSV backend is almost entirely at streaming-level -- unable to accept MapperFeature setting.

What we can do is add something like CsvReadFeature.CASE_INSENSITIVE_HEADERS to allow case-insensitive handling at this level too. But being addition to API, needs to go in 3.2.0.

In 3.1.x it may be necessary to disable CsvReadFeature.FAIL_ON_MISSING_HEADER_COLUMNS in this case, unfortunately.

[cowtowncoder]

Added CsvReadFeature.CASE_INSENSITIVE_HEADERS via #658, will be included in 3.2.0.